on 2/10/2024 8:46 AM, Dave Chinner wrote:
> On Fri, Feb 09, 2024 at 01:20:21AM +0800, Kemeng Shi wrote:
>> I_DIRTY_ALL consists of I_DIRTY_TIME and I_DIRTY, so I_DIRTY_TIME must
>> be set when any bit of I_DIRTY_ALL is set but I_DIRTY is not set.
>>
>> Signed-off-by: Kemeng Shi <shikemeng@xxxxxxxxxxxxxxx>
>> ---
>> fs/fs-writeback.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
>> index 2619f74ced70..b61bf2075931 100644
>> --- a/fs/fs-writeback.c
>> +++ b/fs/fs-writeback.c
>> @@ -1788,7 +1788,7 @@ static int writeback_single_inode(struct inode *inode,
>> else if (!(inode->i_state & I_SYNC_QUEUED)) {
>> if ((inode->i_state & I_DIRTY))
>> redirty_tail_locked(inode, wb);
>> - else if (inode->i_state & I_DIRTY_TIME) {
>> + else {
>> inode->dirtied_when = jiffies;
>> inode_io_list_move_locked(inode,
>> wb,
>
> NAK.
>
> The code is correct and the behaviour that is intended it obvious
> from the code as it stands.
>
> It is -incorrect- to move any inode that does not have I_DIRTY_TIME
> to the wb->b_dirty_time list. By removing the I_DIRTY_TIME guard
> from this code, you are leaving a nasty, undocumented logic trap in
> the code that somebody is guaranteed to trip over into in the
> future. That's making the code worse, not better....
Sure, I will remove this one in next version. Thanks for the
explanation.
>
> -Dave.
>
